Dashboard Week Day 1 - Proper Planning Prevents Piss Poor Performance

by George Walker

The 6Ps above are words to live by, especially in dashboard week. Unfortunately I (re)learned that the hard way today.

Day 1’s task was to look at the the American Housing Survey 2015

Like many census style datasets this was very detailed and had an enormous amount of fields (down to if every room has working electrical plug (2.4% of respondents didn’t). It was also all coded and so required joining onto a helper file – The first challenge for me was finding the correct file to join. There are a number of them on the AHS survey website. Eventually I ended up using the Value Labels Package to create human readable values and then use the codebook to rename my fields.

Once I figured this out it proved reasonably straightforward, but there lay my pitfall. In my haste to get all the data usable I handed considered what variables to transpose and which to keep as key fields, barring a UID.  (see below for my alteryx workflow). This would prove to be an absolute pain in the arse once I actually got into Tableau and started figuring out what I wanted to do with the data

Once I threw the my outputs into tableau I found a couple of interesting things, but found it incredibly difficult to provide any context or interactivity with them because of my earlier foolhardiness with my data prep. This meant either producing a substandard viz, or starting again. Reader, I chose to do both.

Below is my initial “dashboard” exploring the answers to questions about art and cultural events. While it is an OK chart, its nothing to write home about and I wasn’t very happy at all with the outcome.  So like a contestant on Bake Off binning their spongecake and starting again, I went back to data prep, this time with a much clearer idea of what I wanted to do……

Dashboard 1

My first dashboard attempt

So armed with my interesting demographic fields I set about creating a new dashboard. I wasn’t helped by my Tableau crashing and losing a good hour of work, but eventually I produced the below. It isn’t much of an improvement on my previous visualization, but it is about a topic that is more interesting to me than the arts one I plucked out of the data.

Next time (tomorrow) I know I need to take much more time to plan before diving in.

© 2022 The Information Lab Ltd. All rights reserved.